分散式定位於智慧家庭系統中的實作及應用

Abstract

智慧家庭並不是一個新的應用。實際上，在過去二十年間，已經有許多應用被設計並實作出來。然而，一直到目前為止，這樣子的應用並未被廣泛地進入到我們的生活中。其中一個很重要的障礙是，絕大多數智慧家庭系統都需要由專業工程師來做安裝及設定。或許在新建的大樓裡，建商可能會先將系統安裝佈建好，但實際上，即使建商這麼做了，經過一段時間之後，使用者的需求可能會因家庭成員變更等原因而有所改變。另一方面，使用電池作為電力來源的裝置也可能因為電池耗盡而需要做更換，此時，這些修改設定或是重新佈建新的裝置，對使用者來說都將是一件十分浪費時間以及成本高昂的事情。也因此，智慧家庭系統一直沒有辦法在推廣上有長足性的進展。 為了解決這樣的問題，在本篇論文中，我們提出了一個零配置系統(zero-configuration system)的架構。這個系統以自動拓樸重建機制(automatic topology reconstruction mechanism)作為特色之一。在該機制中，我們引入了多維度測距演算法(multidimentional scaling algorithm)來重建複數個裝置的拓樸。藉由拓樸的重建，使用者便能夠輕易地藉由我們設計的人機介面來做裝置的組態控制。此外，為了要消除在多維度測距演算法裡所隱含的旋轉及鏡射問題，我們提出了一個加強型的室內定位演算法來彌補現存之室內定位演算法之精準度不足的問題。此演算法藉由結合地理位置的資訊與距離的資訊來達到提升精準度的目的。透過數值分析的方式可以發現，我們的演算法能夠在最小均方差(minimum mean square error)此一比較基準下提供給現存之室內定位演算法最高80%的加強幅度。

Smart house is not a new research topic. In fact, this topic have been discussed for the past two decades while various applications have been proposed and implemented. However, until now, mass adoption of smart-house applications has not yet happened. One of the main obstacles is that the existing smart houses require professional installation, configuration, and maintenance. Professional installation is not an issue for newly-built houses since devices may be installed and configured by the contractor in advance. However, users' preferences and demands may vary with time due to the change of family composition. Moreover, battery devices such as sensors eventually will run out of power or become malfunctioned so that replacement or rewiring is unavoidable. This process is usually time-consuming and expensive, and impedes the sustainable growth of smart house. In order to solve the problem, a zero-configuration system is proposed in this thesis. The system features an automatic topology reconstruction mechanism. A multidimensional scaling (MDS) algorithm is adopted for implementing the topology reconstruction mechanism. With the reconstructed devices' topology, users are then able to recognize and configure their devices via an intuitive human-machine interface (HMI). To eliminate the mirror and rotation effects introduced by the MDS algorithm, an enhanced indoor localization scheme is also proposed. This localization scheme fuses the geo-coordinates information and distance information to provide more accurate position estimation. Through the numerical analysis, we have shown that the enhanced indoor localization scheme can provide up to 80% improvement on the minimum mean square error (MMSE).